Abstract: Clustering text data streams is an important issue indata mining community and has a number of applications such as newsgroup filtering, text crawling, document organization and topicdetection and tracing etc. However, most methods are similarity-basedapproaches and only use the TF*IDF scheme to represent the semantics oftext data and often lead to poor clustering quality. Recently,researchers argue that semantic smoothing model is more efficient thanthe existing TF*IDF scheme for improving text clustering quality.However, the existing semantic smoothing model is not suitable fordynamic text data context. In this paper, we extend the semanticsmoothing model into text data streams context firstly. Based on theextended model, we then present two online clustering algorithms OCTSand OCTSM for the clustering of massive text data streams. In bothalgorithms, we also present a new cluster statistics structure namedcluster profile which can capture the semantics of text data streamsdynamically and at the same time speed up the clustering process. Someefficient implementations for our algorithms are also given. Finally,we present a series of experimental results illustrating theeffectiveness of our technique.